The Research Of Face Detection Technology For Smart Household Security System

With the rapid development of embedded platform, some visitorauthentication technologies requiring highly computing-performance platformbegan to appear on home security systems. Face detection is a key technology todetect one of the objectives and retrieval technology, and is the premise andfoundation of a number of security technologies. It is very consistent with thecharacteristics of home security environment. Among a variety of biometricbased detection technologies, face detection technology is one of the mostextensive prospects means of identification. Achieve better results with theperformance of face detection is crucial for the security system.Firstly, after analyzing several commonly used face detection design method,the paper presents a face detection algorithm based on the needs and acharacteristic of the subject home security environment and is suitable forembedded home security system. Then the paper uses a simple and efficient Haarfeatures to quantify the value of the image data, and uses of integral imagemethod for rapid calculation of Haar features. All the classifiers used in the papercontain the weak classifiers, the strong classifier, and the cascade classifier. Thehierarchical classification method is used for AdaBoost classifier training.Upon the completion of the face detection algorithm design, the paperdescribes a embedded security system which has a S3C6410processor and runsLinux, and introduces the pre-preparation and construction of the platform forthe development work needed to be done, such as selecting the software andhardware platforms, installing the crossing development tools, and the transplantand configuration process of MJPG-streamer video servers and the OpenCVlibrary. Then this paper focuses on an embedded platform face detection program.The whole face detection program is divided into three parts: the acquisitionprograms, the training programs and the testing procedures. Each section isdescribed in detail about the key code and the running process.After the introduction to the implementation of the face detection system,we conduct a number of experiments and tests. At first we design an experimentfor testing the normal function of face detection. Then we take three sets of testsfor the three important parameters of the classifier such as the number of trainingsamples for face detection algorithm, the proportion of the positive trainingsamples, and the number of weak classifiers of each level strong classifier have.Subsequently, the paper is designed to explore the human face detection system results in practical application scenarios, the paper designs the training set andtesting set contains elevation of the human face, and on this basis a comparisontest. Finally, the characteristics of embedded computing platforms andapplication requirements, the paper also carries a corresponding programcode-level improvements, including a fixed method of floating-point turn,look-up table method, Haar number of features optimized cascade classifiertraining process optimization a variety of methods, also carried out a number oftests, and the experimental results and conclusions.